Blur is much too high in trunk svg filters at the moment : a template problem ?

[ivan louette]

Fortunately I tried to change Inkscape UI language today from my French mother language to English (en). The result was that in svg filters the blur was much too high. However XML values seems to be correct.

I note that if I replace the trunk template with the 0.91 release template the problem is solved.

[su_v (suv-lp)]

> (…) in svg filters the blur was much too high (…)

Would you mind providing the 'steps to reproduce' for a simple but consistent test case? It would ease further investigation and bug triage a lot ... Is this about using the blur slider in Fill & Stroke, about using the blur slider for the feGaussian filter primitive added to a new filter effect in the filter editor, or about blurs in predefined effects, or about blurs in CPFs?

Ideally, test cases attached as saved Inkscape SVG files often are helpful too, in addition to the 'steps to reproduce'.

Do you have any custom (maybe even locale-specific) templates installed in ~/.config/inkscape/templates?

[Alvin Penner (apenner)]

- use 0.91 to create a blur as below, using the default template which has px as the document unit
- use trunk to create a blur as below, using the default template which has mm as the document unit

- create a rectangle of size 50x50 px, stroke width = 1 px
- open Filter Editor, create a new Gaussian Blur filter with a standard deviation of 1.
- note that it is more blurred than the one fom 0.91
- it looks as if the recent trunk version is interpreting the blur standard deviation as 1 mm instead of 1 px.

[ivan louette (ivan-louette)]

I attached two svg files with embedded images and the "Drawing" filter applied

Drawing091 was made with the default.fr.svg template found in usr/share/inkscape/templates on the kubuntu 14.04 partition where my 0.91 is installed

Drawingtrunk is made with the default.fr.svg template found in usr/share/inkscape/templates on the kubuntu 14.04 partition where my trunk is installed

If I copy and paste the image with wrong one into the file with the correct one the wrong remains wrong and the correct remains correct.

[Alvin Penner (apenner)]

it is possible that the issue I was referring to in the two files rect_trunk13954.svg and rect_0.91_stable.svg may not be the same issue as the isuue raised by ivan, so I better clarify why I objected to this behaviour. I don't object to the fact that 0.91 interprets the blur width as px while the trunk interprets the blur width as mm. That behaviour is perfectly consistent with the document-units that are used in the two cases. What I object to is the fact that the Filter Editor does not explicitly state what the units are. It presents the Standard Deviation as a variable with no units, which of course is not true, it has units which may be either px or mm depending on what the document-units are. This should be stated explicitly in the Filter Dialog, possibly with the option of changing the units if desired.

[ivan louette (ivan-louette)]

Now the same problem occurs with Turbulence. I attach a file with two versions of the same image. The right one is made with 0.91 template and the the left one with trunk template

[ivan louette (ivan-louette)]

The bug I describe is exactly what Alvin tells. And of course a workaround is easy : simply changing the users unit to 3,7775 by mm in the Trunk solve it.

However users will be surprized by the big differences while applying the same filters with the same settings in 0.92 if the question isn't solved.

Another question is : is there a default size for standard blur deviation in SVG specifications ? If this is the case then we should of course apply it.

[ivan louette (ivan-louette)]

To develop Alvin's idea what could be interesting is adding the ability to change Filter effects/Units in Preference/Rendering, and adding the possibility to numerically change the Deviation Scaling Factor in Filter Editor/Filter General Setting tab. This interactivity would be very useful in many cases.

[ivan louette (ivan-louette)]

Sorry : I wrongly wrote 3,7775 by mm in #9 instead of 3,7795